From keywords to discursive legitimation: Representing 'kuffar' in jihadist propaganda magazines Anina L. Kinzel College of Arts and Humanities Swansea University Singleton Park Swansea, Wales SA2 8PP [email protected] Abstract had to be written in English and meet the project’s definition of a magazine (Macdonald et al., 2015). This paper explores the integration of Corpus The reason this paper focuses on Dabiq and Linguistics and Discourse Analysis Inspire, produced by IS and Al-Qaeda, methodologies, specifically the use of keyword respectively, is that these two groups currently analysis to pinpoint salient discourse pose the biggest terrorist threat to the Western representations of given social identities. The world. Terrorist groups are also often regarded as particular identity examined here is that of ‘kuffar’, which is a derogatory term used by acting the same, especially if they follow a jihadist jihadist-ideology groups to describe people who ideology. However, they might target different do not share their faith. Two measures (Log social groups and compete for recruits in fighting Likelihood and %DIFF) are used to determine a common enemy – usually an out-group who the keyness value of the term kuffar in the does not share their ideology and faith. propaganda magazines produced by two such Comparative research within these groups is groups: Al Qaeda (Inspire) and Islamic State therefore most useful to understand their (Dabiq). Although they yield different keyness similarities and differences. The aim of this paper values, both confirm the salience of this social is twofold: (1) to contribute to the current identity within jihadist propaganda. The results academic debate about one of the main analytic of a Key Word in Context concordance analysis of the term kuffar show how negative tools in Corpus Assisted Discourse Studies representations thereof are mainly legitimized (CADS): keywords; and (2) to show how IS and on impersonal authority grounds (Van Leeuwen, Al-Qaeda discursively construct “kuffar” in their 2007). The results also reveal a trend towards propaganda magazines and legitimize this as a reminding Muslims of their “duty” to fight and negative social identity. “Kuffar” is a derogatory kill kuffar individuals. This is supported by term, which describes individuals on religious positive expressions to describe Muslims who grounds, or lack thereof. Analyzing this term perform this duty, and negative expressions and provides insights into the role of religion in traits associated with the non-believers. This jihadist ideologies, which is still heavily debated. case study supports the view that CL and Discourse Analysis can offer a ‘useful synergy’ (Baker et al. 2008, 2015) to research into, amongst other areas, representation and 1.1 CADS ideology in language. Corpus-Assisted Discourse Studies (CADS) is the application of Corpus Linguistics (CL) tools such 1. Introduction as keywords, wordlist, and concordance line analysis in Discourse Studies. The field has been The research in this paper was carried out as part evolving since the 1990s, when new Corpus of a research project funded by Swansea Linguistics tools were developed. Until 2004 not University and directed by professors Stuart many studies used a CADS approach (Partington, Macdonald and Nuria Lorenzo-Dus. The project 2004). Partington describes CADS as ‘the examined jihadist propaganda from different uncovering, in the discourse type under study, of disciplinary perspectives, including Terrorism what we might call non-obvious meaning’ Studies and Linguistics. All available online (Partington, 2008, p. 191). It combines the publications of five jihadist magazines released statistical techniques of CL, such as keyword and between January 2009 and the end of June 2015 frequency lists and concordancing, with the were collected. These publications are Jihad qualitative tools of Discourse Analysis, namely Recollections, Gaidi Mtaani, Inspire, Dabiq and close-readings and reflection (Partington, 2008) to Azan. To be included in the data set, publications uncover non-obvious meaning. Baker 2015 argues 26 that using a CADS approach, especially with a two wordlists of the two corpora. Words with a large corpus, removes some of the bias a DA high ‘keyness’ appear in the given text more researcher may have. It also gives DA researchers frequently than expected. Although there are a starting point: a CADS methodology may drive different statistical techniques to determine the the analysis in ‘unexpected directions’ (Baker, ‘keyness’ of words, log-likelihood (Baker et al 2015, p. 144). In a 2008 journal article, Baker et 2006b) is arguably the standard one. It is also a al. conclude that using Corpus Linguistics tool technique that has received some criticism, and Critical Discourse Analysis is ‘a useful leading to the CL / CADS research community methodological synergy’ (Baker, et al., 2008, p. seeking to develop alternative techniques. One 273). Some examples of work that uses a CADS such technique, which is compared in this paper approach to examine how Muslims are to Log Likelihood, is called %DIFF. Introduced constructed by various media are as follows: by Gabrielatos and Marchi (2011, 2012), its main Baker et al. use Sketch Engine, a corpus analysis difference lies in that it is based on effect size tool for grammatically tagging items, and other CL techniques to examine how collocates of the 1.2.1 Log-Likelihood word ‘Muslim’ is constructed in British Newspapers articles between 1998 and 2008 Log-likelihood is a test that shows statistical (Baker et al., 2013). In their analysis they focus on significance (Baker et al., 2006b). It assigns a p- the two most frequent immediate right-hand value to every word and measures how significant collocates “world” and “community” (Baker et the word is in comparison to a reference corpus. al., 2013). Their findings include that Muslims are According to Biber et al. the p-value ‘represents often associated with negative aspects, they are the probability that this keyness is accidental’ also portrayed as being easily offended and in (Biber, et al., 2007, p. 138). Log-Likelihood does conflict with non-Muslim communities (Baker et not take into consideration the size of the study al., 2013). Muslim world was often referring to corpus, which means it does not reflect the different branches of Islam as one while being definition of keywords (Gabrielatos and Marchi, constructed on language or geographical grounds, 2011). If an English language magazine, which rather than religious grounds (Baker et al., 2013). also features some Arabic words was compared to McEnery et al. also look at the word “Muslim” as an English language corpus, such as the British part of their analysis of media reactions to the National Corpus (BNC) (British National Corpus, murder of Private Lee Rigby by Michael 2007), all Arabic words would be assigned a low Adebolajo and Michael Adebowale, who p –value, because they appear much more converted to Islam (McEnery et al., 2015). The frequently in this magazine. However, this does findings showed that the word is associated with not mean that all of these words are especially key the murderers and the victim, but is used to in this corpus or indicate aboutness, these words distance this action from other Muslims. Islam as are just different compared to the BNC. a keyword is associated with negative aspects 1.2.2 Effect size (%DIFF) such as betrayal, radicalization and extremism (McEnery et al., 2015). Effect size measures ‘the difference of the 1.2 Keyword analysis frequency of a word in a study corpus when compared to that in the reference corpus’ Keyword analysis is a much used technique of CL (Gabrielatos and Marchi, 2011). Contrasting to (Gabrielatos and Marchi, 2011). Bondi notes that log-likelihood, the size of the two corpora is also ‘[T]he study of keywords has become central in taken into account, which is especially useful corpus linguistics, especially through the when comparing corpora of different sizes. In this development of techniques for the analysis of the case, the reference corpus does not have to be meaning of words in context’ (Bondi, 2010, p. 3). larger than the study corpus. In the broadest sense, keywords are words that are 1.3 Structure important in a given text (Stubbs, 2010). They mirror what the text is about (Scott and Tribble, This paper will first compare the results of a 2006), which is why they are an important tool of keyword analysis of the corpus (see 3.1), focusing CL to help identify a subset of textual items to analyze (Baker, 2006a). Keyword lists in CL tools such as AntConc (Anthony, 2016) and WordSmith (Scott, 2016a) are based on statistical significance compared to a usually larger reference corpus. This is done by first compiling 27 on the term ‘kuffar’1, by applying log-likelihood could not be assigned to any legitimation and %DIFF measures. It will then report the category: they were general instances of “kuffar” results of a KWIC analysis for “kuffar” that shows in which the derogatory term was not used how its negative discourse representation is alongside any legitimating behavior, such as legitimized by the magazine authors. rulings on fighting against and killing “kuffar”. 3. Methodology The corpus was not lemmatized because 3.1 Data one of the project’s aims was to examine stylistic variation, including at the spelling level, between The corpus for this study consist of 22 issues of the different magazines, especially regarding use Dabiq (185,951 words) and Inspire (304, 347 of Arabic terms. words). 13 of these issues are from Inspire and were published between the summer of 2010 and 4. Results the spring of 2015. The remaining nine issues were published by IS between the summer of 2014 In CADS, the analysis usually starts with a and the spring of 2015. The digital versions of the keyword analysis. As can be seen in Tables 1 and issues were collected in the summer of 2015, 2, the log-likelihood for “kuffār” is the same in coinciding with the onset of the project.